The　performance　of　XPath　query　is　the　key　factor　to　the　capacity　of　XML　processing.　It　is　an　important　way　to　improve　the　performance　of　XPath　by　making　full　use　of　multi-threaded　computing　resources　for　parallel　processing.　However,　in　the　process　of　XPath　parallelization,　load　imbalance　and　thread　inefficiency　often　lead　to　the　decline　of　parallel　performance.　In　this　paper,　we　propose　a　cost　optimization-based　parallel　XPath　query　method　named　coPXQ.　This　method　improves　the　parallel　processing　effect　of　navigational　XPath　query　through　a　series　of　optimization　measures.　The　main　measures　include　as　follows:　first,　by　optimizing　the　storage　of　XML　node　relation　index,　both　storage　and　access　efficiency　of　the　index　are　improved.　Secondly,　load　balancing　is　realized　by　a　new　cost　estimation　method　according　to　the　number　of　XML　node　relations　to　optimize　parallel　relation　index　creation　and　parallel　primitive　execution.　Thirdly,　the　strategy　of　determining　the　number　of　worker　threads　based　on　parallel　effectiveness　estimation　is　utilized　to　ensure　the　effective　use　of　threads　in　query.　Compared　with　the　existing　typical　methods,　the　experimental　results　show　that　our　method　can　obtain　better　parallel　performance.